Facial recognition from CCTV
Data Analytics Platform
The project came to inception during HackTrain 3.0, with the aim of creating a product which tackles customer service in the rail industry. Tom Ewing (Data Scientist @ TfL) proposed the idea in our team, and with discussions with the CTO of Eurostar and representatives from SNCF, we took feedback during the design process, to nail down the development plan.
What it does
By using OpenCV and python, the footage from CCTV is taken and faces in the trains are analysed for different emotions. If anger is detected as an average metric over a period of 30 seconds or more, an attendant is notified through a mobile app and is then told the seat number. Anonymous data is then collected, so train companies are able to collect information, and have another metric to use when trialling different customer service strategies.
How I built it
There were three parts of the system: Sense, Respond and Know
Sense By using Python and OpenCV, we were able to extract faces from frames in the footage, and have these saved temporarily. After collecting several, these were sent off in a batch to a Microsoft API, and the results were then returned.
Respond If a customer is unhappy for a period of time, a notification is sent to a mobile app, which is used by operators. They then know the seat number and booking information of the passenger, and can provide personalised service.
Know We then collected the information, and began to build up a data set, with the hope that meaningful patterns could be found in the future.
Accomplishments that I'm proud of
Awarded SNCF's "Most Innovative Product" for HackTrain 3.0
What's next for eMotive
The product will be pitched to leaders of the Rail Industry on the 15th of November at Runway East, London.